Factors affecting the efficiency of the BRICSs' national innovation systems: A comparative study based on DEA and Panel Data Analysis

Cai, Yuezhou Working Paper Factors affecting the efficiency of the BRICSs' national innovation systems: A comparative study based on DEA and Panel Data Analysis Economics Discussion Papers, No. 2011-52 Provided in Cooperation with: Kiel Institute for the World Economy – Leibniz Center for Research on Global Economic Challenges Suggested Citation: Cai, Yuezhou (2011) : Factors affecting the efficiency of the BRICSs' national innovation systems: A comparative study based on DEA and Panel Data Analysis, Economics Discussion Papers, No. 2011-52, Kiel Institute for the World Economy (IfW), Kiel This Version is available at: https://hdl.handle.net/10419/52679 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by-nc/2.0/de/deed.en Factors Affecting the Efficiency of the BRICSs’ National Innovation Systems: A Comparative Study based on DEA and Panel Data Analysis Yuezhou Cai Chinese Academy of Social Sciences (CASS) Abstract Efficiency scores for the national innovation systems (NISs) in 22 countries including the BRICS and the G7 are calculated using data envelopment analysis (DEA). Factors that may affect the efficiency of the NIS are summarized based on the NIS approach and the new growth theory. Empirical evidence is provided using panel data analysis and principal component analysis. The results of the efficiency calculations and the empirical tests show the following. (1) The BRICS differ greatly in the efficiency of their NISs, with China, India, and Russia ranking fairly high, and Brazil and South Africa ranking low. (2) In accordance with the NIS approach and the new growth theory, there are many factors that affect the NISs including ICT infrastructure, enterprise R&D, market environment, governance, education systems, economic scale, natural endowments, and external dependence. (3) Enterprise innovation is of particular importance for the NISs. To improve the efficiency of innovation systems, efforts should be made to improve the market conditions, governance, and financial structures, and create a sound environment for R&D. (4) ICT infrastructure, economic scale, and openness affect the diffusion of knowledge and technology, and in turn affect NIS efficiency. (5) The BRICS have low governance levels and a high dependency on natural resources, both of which are determined by their stage of development and extensive growth patterns. To avoid the socalled middle-income trap, the BRICS should transform their factor-driven growth patterns into innovation-driven growth patterns. China still needs to improve its ICT infrastructure, its governance systems, and its education system. During its 12th five-year plan, more effort should be devoted to these fields and to improving external conditions for R&D. JEL O30, O57, P52 Keywords The BRICS; National Innovation System (NIS); NIS efficiency; Data Envelopment Analysis (DEA); Panel Data Analysis (PDA) Correspondence Yuezhou Cai, Institute of Quantitative & Technical Economics (IQTE), Chinese Academy of Social Sciences (CASS); e-mail: [email protected]. This paper was first presented at the workshop “Innovation: from Europe to China” held by the Kiel Institute for the World Economy in Kiel, Germany, on Oct. 28-29, 2011. The views in this paper are not the views of the IQTE CASS. The author is grateful to John Whalley at the University of Western Ontario, Zhou Chunyan at the Universidad Complutense de Madrid, Aoife Hanley, Liu Wan-Hsin, Dirk Dohse at the Kiel Institute for the World Economy, Sourafel Girma at Nottingham University, and Wang Tongsan, Li Jinhua, Zhang Tao,and Liu Shenglong at the IQTE CASS, for their comments and discussions. This research was partly subsidized by the Key Subject Construction Project of the IQTE, CASS. © Author(s) 2011. Licensed under a Creative Commons License - Attribution-NonCommercial 2.0 Germany Discussion Paper No. 2011-52 | December 6, 2011 | http://www.economics-ejournal.org/economics/discussionpapers/2011-52 conomics Discussion Paper Introduction As the representatives of emerging economies, during 2000-2009, Brazil, India, Russia, China and South Africa (BRICS) have gained an average annual growth of 3%, 5.4%, 7.2%, 10.3% and 3.6% respectively, much higher than the OECD average of 1.6% and the World average of 2.6%. However, rapid growth does not enhance the competitiveness of the BRICS correspondingly. According to the ranking of Global Competitiveness Report provided by the World Economic Forum, the competitiveness of Brazil, Russia, India, China, and South Africa rank 31, 52, 37, 44, and 25 in the year 2000. By 2009, their ranking changed to be 58, 63, 49, 29, and 45. All but China get a decrease in competitiveness ranking to certain extent. Even for China, the competitiveness ranking of about 30 does not match with its world highest growth rate and second largest economic size. The paradox of high growth and low competitiveness for the BRICS can be attributed to their extensive growth patterns. In recent years, the high growth of China and India is largely dependent on their advantages in demographic structure and cheap labor force, while Russia, Brazil and South Africa depend more on the export of mineral resources. As we know, it is a very common chooses for economies in the take-off stage to choose an extensive growth pattern characterized with high-input, high-accumulation and high-output. However, most of the BRICS have stepped over the stage and entered a transition from middle-income to highincome. The growth of these countries will inevitably encounter constraints such as natural resources supply, environmental degradation, etc. To maintain a rapid growth and surpass "Middle-income Trap", it is very urgent for the BRICS to improve the efficiency of National Innovation System (NIS) and enhance their innovation capability. Therefore, efficiency of NIS and the influencing factors behind should be calculated and analyzed before a series of policies and measures were adopted. As a matter of fact, the NIS of the BRICS has increasingly been studied while their economies are uprising in recent 10 years. Liu and White (2001)[ 14 ]put forward an analytical framework for Innovation System including R&D, implementation, end-use, education, and their linkages, being applied to make a comparative study on China’s NIS in different periods and development stages. Viotti (2002)[ 19 ] proposed a new conceptual and theoretical framework of "National Learning System (NLS)", for developing economies based on the NIS Approach. Viotti (2002) further chose Brazil and South Korea as typical cases, and characterized their NLS as Passive and Active one correspondingly. Feinson (2003)[ 6 ] also chose the two as typical cases but proposed that R&D policies, intellectual property, human capital, FDI are all among the elements affecting the NIS besides NLS. The joint research project co-directed by Cassiolato and Lundvall (2010) [ 2 ]makes overall comparison for the NIS of BRICS in production, education, finance, politics, regulation, etc. www.economics-ejournal.org 3 conomics Discussion Paper Tomas et al (2011)[ 18 ] analyze R&D efficiency of 51 regions in the United States from 2004 to 2008 with the ratio of patents granted and scientific publications to R&D expenditures, and compared the performance of United States with that of the BRICS. Most of the existing literatures on NIS of the BRICS are mainly cases study or qualitative description of the innovation patterns, while the quantitative analysis on innovation capability and efficiency of innovation is very limited. To reduce the subjective factors as far as possible and get a more objective and reliable judgment on the NIS of the countries, the Data Envelopment Analysis (DEA) and econometric analysis are adopted in this paper. Here, the NIS is treated as a special production sector in the economy with certain inputs, including human and financial resources to produce some particular outputs such as patents, scientific publications. The NIS of the BRICS as well as other 17 nations including the G7 and some other OECD members are chosen as the Decision Making Units (DMUs). Relative efficiency scores of each DMUs are calculated with DEA method. Factors influencing NIS efficiency are summarized according to the National Innovation System Approach (NIS Approach) and New Growth Theory. And empirical test is carried out using cross country panel data with the efficiency scores as the explained variable. The NIS efficiency as well as the influencing factors of China and other four countries will be further analyzed based on the previous DEA efficiency scores and the panel data analysis. Section 1 is a brief review on quantitative methods for analysis of NIS. Section 2 is the relative efficiency calculation of NIS for 22 countries including the BRICS. Section 3 will analyzes the factors affecting the NIS efficiency. And in the end some concluding remarks and policy implications are explored in section 4. 1 Quantitative methods for analysis of NIS When Christopher Freeman, Bengt-Ake Lundvall, Richard Nelson and other innovation economists gave the concept of National Innovation System (NIS) and the NIS Approach in the late 1980s and early 1990s, the research methods in this field are mainly qualitative analysis such as evolutionary analysis and case study. In the mid of 1990s, many innovation economists proposed to adopt more quantitative methods in the study. Patel and Pavitt (1994)[ 17 ] might first appeal for quantitative analysis on the input and output characters of NIS 1 .With the execution of Community Innovation Survey (CIS) throughout the European Union and its implementation in the annual European Innovation Scoreboard 1 The quantitative analysis in the field of innovation can be further track to 1950s and 160s, when the Linear Model of Innovation was prevailing. www.economics-ejournal.org 4 conomics Discussion Paper (EIS) 2 ,quantitative analysis has been widely used in studying national innovation capacity and innovation system. These quantitative methods can be categorized into 3 approaches: Composite (Innovation) Indicators, DEA Efficiency Calculation, and Modeling/Econometric Approach. Composite Indicators Approach has been adopted by many institutions in evaluating innovation capacity in national level. The famous EIS/IUS, Competitiveness Ranking of the World Economic Forum, and many other ranking works are all implementation of the Approach. To get a sound and comparable evaluation results under this approach, an inclusive indicator system covering various aspects of NIS is to be established. Normally, the indicator system would include indicators: input, output, procedure of innovation; as well as organization pattern of innovation activities, institutional arrangements, etc. However, here efficiency of innovation system is ignored since the input and output indicators are treated in the same way. As a result, economies with "high innovation inputs and low innovation outputs" may get a score equal to or even higher than those with "low innovation inputs and high innovation outputs". In contrast to the composite Indicator approach, the DEA Approach focused exactly on input-output efficiency of innovation systems. In this approach, innovation system is treated as a special sector of the economy, and each (country or region) economy is regarded as an independent DMU (Decision Making Unit). After choosing proper innovation input and output indicators, the relative efficiency of each DMU can be calculated. Nasierowski and Arcelus (2003) [ 15 ] has applied the DEA to calculate efficiency of innovation system for over 40 countries and regions. Guan Jiancheng et al (2006) [ 10 ] also have done a lot of innovation efficiency calculation with DEA in both country and regional levels. An obvious advantage of DEA Approach is the simplification of indicator system since only indicators of innovation inputs and outputs are required. At the same time, the efficiency score calculated from these input/output indicators is a reflection of the capability of transferring innovation inputs into outputs, and can be regarded as a composite capability of the Innovation System as well. Nevertheless, what the efficiency score reflects is only the general capability of the Innovation System, while the information of what influence this capability are not given by the scores. The Modeling/Econometric Approach is mainly used to analyze the factors influencing the national innovation capacity. The procedure of this approach includes theoretical analysis, mathematical modeling, and econometric test, which is conformed to the research paradigm of mainstream economics. Furman et al (2002, 2004)[ 9 ] [ 8 ], Hu and Mathews (2005, 2008) [ 12 ] [ 13 ]also explored the approach. Factors analysis in this approach is supported by both economic theory 2 In October, 2010, the former EIS was reconstructed and renamed as Innovation Union Scoreboard (IUS). www.economics-ejournal.org 5 conomics Discussion Paper and the empirical data, with more reliable results. However, in econometric test, one single indicator is chosen as the explained variable, for example, "International Patent Granted" is usually selected as the proxy of innovation capacity. As we know, innovation capacity is far from the patents granted. And such treatment would inevitably lead to a bias. Obviously, the advantages of above two approaches, are complementary, and can be combined into a deep and overall analysis on efficiency of NIS. Actually, the combination has been applied in the empirical study of other fields. For example, Casu and Molyneux (2003) [ 3 ]calculated the relative efficiency scores for 530 European Banks. Taking the efficiency scores as the explained variable, they then analyze the factors affecting the efficiency, with Tobit model. Hoff (2007)[ 11 ] named the DEA-based econometric analysis as "Second stage DEA" and compares the estimation results between OLS and Tobit. It is shown that OLS may actually in many cases replace Tobit as a sufficient second stage DEA model. In the empirical studies of NIS efficiency, few have combined the DEA with econometric analysis. This paper will fill the gap to make such a combination for analyzing NIS efficiency of the BRICS. 2 Measuring the relative efficiency of NIS with DEA 2.1 Optimization model for the efficiency measurement The DEA method has been used in measuring the relative efficiency for DMUs widely, since it was first proposed by Charnes et al (1978)[ 4 ] over 30 years ago. According to Farrell (1957)[ 5 ], the measurement of production efficiency can be divided into two categories: Input-Oriented and Output-Oriented. The former fixes outputs to compare the inputs to measure relative efficiency, while the later fixes inputs to compare the outputs. Output-Oriented model is chose in this paper. Besides, efficiency measuring models can be divided into Constant Return to Scale (CRS) and Variable Return to Scale (VRS). More DMUs would be in the production frontier using VRS model compared with using CRS model, which means more DMUs would get an efficiency score of 1. Such a measuring result may lead to a bias in the efficiency scores and in turn affect a succeeding econometric analysis. So, the CRS model is used in the final measurement to reduce the above bias. And the algebraic expression for the chosen Output-Oriented CRS model is as follow. (Dt o)          [Dt o(−→ x0t,−→ y0t)]−1=Maxφ,λφ s.t.−Xt m×n −→ λn×1+−→ x0t≥0m×1 Yt s×n −→ λn×1+0s×1φ≥φ−→ y0t φ>0,λj≥0,j=1,2,···,n (1) www.economics-ejournal.org 6 conomics Discussion Paper In the above equation (1), X , Y are the inputs and outputs matrixes formed by of all DMUs; n,mand srefer to the number of DMU, input indicators and output indicators respectively;While t and φ refer to time period and relative efficiency score. What can be calculated from equation (1) is the relative efficiency score of DMU0 in the period of t . And the relative efficiency score φ is also the value of distance function to the Frontier for the DMU being measured, which can be expressed as [Dt o(−→ x0t,−→ y0t)]−1.3 2.2 Selection of DMUs and input/output indicators The efficiency score is highly relevant to the input/output indicators as well as the number of DMUs. If the number of DMUs to be measured is very limited, most DMUs may be in the production frontier constructed by themselves with the corresponding relative efficiency score to be 1. And the ranking based on these efficiency score may become meaningless. To avoid such circumstance, the G7 countries (United States, Japan, Germany, Canada, United Kingdom, France, Italy), 8 European countries (Finland, Sweden, Denmark, Swiss, Netherland, Austria, Belgium), and two OECD countries in the Asia Pacific Region, (South Korea and Australia) are selected as the DMUs, together with the five BRICS countries. The 17 countries are all OECD members, including the world largest developed economies, small European economies famous for their innovation capacity and competitiveness, and South Korea, the typical successful case of catching up with and surpassing. Taking the NIS as a special production sector, the production frontier constructed by the above countries (DMUs) would definitely be a good approximation of the real one. As a special production sector, the inputs of NIS are mainly the human resources and financial resources allocated in innovation activities, while the outputs are mainly patents granted, scientific publications, and output of high-tech industries. Relevant input-output indicators from 2000 to 2008 are collected from different sources including the World Bank’s Open Database (DataBank), the database of UNESCO Institute for Statistics, and the data released in the website of World Intellectual Property Organization (WIPO). Finally, "General Expenditures on R&D (GERD)" and "Total R&D personnel" are chosen as the input indicators, while the output indicators include "WIPO patents granted", "Scientific and technical journal articles", and "High-technology and ICT services exports". Although there is no dimensional limit using DEA for efficiency measurement, the indicators of GERD and "High-technology and ICT exports" are still converted into constant price 3 The index of technological change for the DMU, that is the Malmquist Index, can be further calculated based on equation (1) via geometric average. www.economics-ejournal.org 7 conomics Discussion Paper of 2000 U.S. dollar, 4 and the "Total R&D personnel" is converted into full-time equivalent (FTE). 2.3 Measurement results of the NIS efficiency score Based on the previous measuring model and the data collected, the relative efficiency scores for the NIS of the 22 countries from 2000 to 2008 are calculated with Win4DEAP. See Table 1 and Table 2. Table 1: Relative efficiency scores for NIS of 22 countries including BRICS 2000 2001 2002 2003 2004 2005 2006 2007 2008 BR 0.184 0.223 0.233 0.198 0.214 0.219 0.223 0.198 0.213 RU 0.756 0.680 0.773 0.676 0.628 0.756 0.727 0.601 0.748 IN 0.415 0.588 0.680 0.642 0.661 0.675 0.735 0.756 0.913 CN 0.631 0.731 0.849 0.870 0.920 0.926 0.956 0.944 0.999 ZA 0.314 0.343 0.311 0.274 0.264 0.265 0.266 0.250 0.261 US 0.340 0.353 0.352 0.251 0.250 0.254 0.274 0.249 0.264 JP 0.358 0.357 0.385 0.308 0.308 0.320 0.336 0.300 0.310 DE 0.368 0.460 0.523 0.429 0.474 0.467 0.499 0.415 0.448 UK 0.818 0.608 0.748 0.595 0.661 0.688 0.773 0.676 0.593 FR 0.362 0.395 0.414 0.323 0.311 0.316 0.347 0.312 0.357 CA 0.455 0.412 0.390 0.332 0.279 0.294 0.318 0.315 0.344 IT 0.367 0.419 0.417 0.376 0.369 0.362 0.364 0.338 0.357 FI 0.436 0.442 0.487 0.403 0.353 0.423 0.386 0.378 0.546 SE 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 DK 0.353 0.395 0.473 0.369 0.361 0.385 0.351 0.296 0.281 CH 0.554 0.697 0.859 0.725 0.738 0.692 0.725 0.686 0.842 NL 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 NO 0.235 0.296 0.333 0.230 0.220 0.242 0.279 0.254 0.335 AT 0.446 0.517 0.658 0.526 0.558 0.430 0.454 0.390 0.404 BE 0.511 0.620 0.829 0.680 0.681 0.678 0.639 0.621 0.732 AU 0.226 0.244 0.243 0.214 0.192 0.244 0.178 0.159 0.162 KR 0.832 0.682 0.849 0.702 0.704 0.682 0.989 0.890 0.737 Means 0.498 0.521 0.582 0.506 0.507 0.521 0.537 0.501 0.538 Notes: each country is abbreviated as follow, BrazilBR, Russia FederationRU, IndiaIN, ChinaCN, South AfricaZA, United StatesUS, JapanJP GermanyDE, United KingdomUK, FranceFR, CanadaCA, ItalyIT, FinlandFI, SwedenSE, DenmarkDK, SwitzerlandCH, NetherlandsNL, NorwayNO, AustriaAT, BelgiumBE, AustraliaAU, KoreaKR. 4 Converting into constant price of 2000 U.S. dollar is convenient for further calculation of Malmquist Index in this period. www.economics-ejournal.org 8 conomics Discussion Paper Table 2: Efficiency ranking for NIS of 22 countries including BRICS 2000 2001 2002 2003 2004 2005 2006 2007 2008 BR 22 22 22 22 21 21 21 21 21 RU 5 6 7 7 9 5 7 9 6 IN 12 9 9 8 7 9 6 5 4 CN 6 3 4 3 3 3 4 3 3 ZA 19 19 20 18 18 18 20 19 20 US 18 18 18 19 19 19 19 20 19 JP 16 17 17 17 16 15 16 16 17 DE 13 11 11 11 11 10 10 10 11 UK 4 8 8 9 8 7 5 7 9 FR 15 15 15 16 15 16 15 15 13 CA 9 14 16 15 17 17 17 14 15 IT 14 13 14 13 12 14 13 13 14 FI 11 12 12 12 14 12 12 12 10 SE 1 1 1 1 1 1 1 1 1 DK 17 16 13 14 13 13 14 17 18 CH 7 4 3 4 4 6 8 6 5 NL 1 1 1 1 1 1 1 1 1 NO 20 20 19 20 20 20 18 18 16 AT 10 10 10 10 10 11 11 11 12 BE 8 7 6 6 6 8 9 8 8 AU 21 21 21 21 22 22 22 22 22 KR 3 5 5 5 5 4 3 4 7 Means 8-9 9-10 10-11 10-11 10-11 9-10 9-10 9-10 10-11 The above measurement results show that the relative efficiency of the NIS of the BRICS during 2000-2008 differs a lot from each other. In general, Russia, India, and China get pretty good scores and rankings with China ranking 3-4 in most of the period. At the same time, the scores and rankings of Brazil and South Africa are far from satisfactory, almost always ranking at the bottom. The efficiency score and ranking of NIS for G7 countries are not as striking as anticipated, comparing with their leading status in the world economy. Only the United Kingdom gets an efficiency score above mean level for the 9 years. Following the United Kingdom, the efficiency score of Germany is a little bit lower than the mean level for most of the period. In contrast, more small developed European economies get a good efficiency score. Among them, the NIS of Sweden and Netherland are always in the production Frontier with an efficiency score of 1. Swiss, Belgium, and Austria get fairy good scores and high rankings as well. Nevertheless, the efficiency scores of the NIS for the remaining three small economies, Norway, Denmark, and Finland are below the mean level in most of www.economics-ejournal.org 9 conomics Discussion Paper However, there still are some exceptions. The coefficients of CDTPRV and CAPLST change across different models, and the coefficient sign of POLSTAB and VOACCT is not conformed to the economic intuition in spite of their robustness. 9 3.4 Panel data analysis with some variables constructed by principal factor analysis A possible explanation for the above exceptions is that there exists multi-collinearity due to the correlation among the proxy variables representing the same influencing factor. To eliminate the correlation among proxy variables, principal factor analysis may be applied to generate few independent principal factors or just the first principal factor, which contain most of the information carried by the relevant proxy variables. Replacing the proxy variables corresponding to the same influencing factor with their first principal factor, the multi-collinearity among explanatory variables would be eliminated. Next, panel data analysis would be further applied to the efficiency of NIS with some constructed first principal factors as explanatory variables. How the first principal factors were constructed is listed in Table 6. Table 6: Construction of first principal factors for relevant factors First Principal factors (F. P. F.) Constructing equations Cumulative proportion Infrap1(Infrastructure) Infrap1=0.62ITNET+0.56 MOBL+0.55 TEL 0.75 Finp1 (financial structure) Finp1=0.64CDBBAN+0.66 CDTPRV+0.40 CAPLST 0.71 Entp1 (enterprises innovation) Entp1=0.55ENTRRE+ 0.60ENTRPGERD+0.57 ENTRFGERD 0.78 Edup1 (education system) Edup1=0.71 TEENRL+0.71 SEENRL 0.89 Markp1(Market circumstance) Markp1=-0.38 BSCOST-0.60 TAXRATE+0.03 PRCOST+ 0.41 0.60 LGRIGHT+0.36 INVPORT Govp1 (governance) Govp1=0.42CORRUP+0.42GOVEFF+0.39POLSTAB+ 0.92 0.41REGULA+0.41 LAW+ 0.39VOACCT The six first principal factors contained most of the information carried by the corresponding proxy variables for ICT infrastructure, financial structure, enterprise innovation activities, education system, market circumstance, and governance. Panel data analysis might be made once again with the above principal factors and some of the retained indicators, NRTGDP, TRTGDP, RGDPPC, and PORGDP as the explanatory variables. And the regression will also be made under the rule of "General-to-specific" with the insignificant variables being eliminated gradually. The regression results with principal factors as explanatory variables are listed in Table 7. 9 For example, a stable political situation forms a good social environment for innovation, and should enhance the efficiency of the NIS. However, the POLSTAB is negatively related to NIS efficiency according to the regression results in table 5. www.economics-ejournal.org 16 conomics Discussion Paper Table 7: Regressions with First Principal Factors as Explanatory Variables Variables Model 1 Model 2 Model 3 Model 4 Model 5 C -0.083(0.600) -0.081(0.554) -0.083(0.541) -0.038(0.755) 0.010(0.924) INFRAP1? 0.000(0.134) 0.000(0.029) 0.001(0.013) 0.001(0.018) 0.001(0.013) FINP1? -0.001(0.000) -0.001(0.000) -0.001(0.000) -0.001(0.000) -0.001(0.000) ENTP1? 0.000(0.517) 0.000(0.516) 0.000(0.526) – – EDUP1? 0.000(0.978) – – – – MARKP1? -0.005(0.037) -0.005(0.032) -0.005(0.037) -0.005(0.028) -0.006(0.012) GOVP1? -0.007(0.603) -0.007(0.587) – – – RGDPPC? 0.028(0.283) 0.028(0.269) 0.024(0.287) 0.019(0.403) – PORGDP? 0.068(0.000) 0.068(0.000) 0.069(0.000) 0.072(0.000) 0.078(0.000) TRTGDP? 0.002(0.000) 0.002(0.000) 0.002(0.000) 0.002(0.000) 0.002(0.000) NRTGDP? -0.002(0.377) -0.002(0.372) -0.002(0.423) -0.003(0.201) -0.003(0.153) R-Sq 0.99 0.99 0.991 0.991 0.991 Adjusted R-Sq 0.989 0.989 0.989 0.99 0.99 D.W. 1.62 1.62 1.608 1.611 1.615 Cross section F 185.19(0.000) 188.29(0.000) 196.64(0.000) 314.51(0.000) 315.78(0.000) It can be seen from Table 7 that the three constructed first principal factors, INFRAP1, FINP1, and MARKP1, are statistically significant. The coefficients of these variables show good robustness in different models, and the sign of the coefficients are conformed to economic intuitions on the whole. The rest 3 principal factors, ENTP1, EDUP1 and GOVP1 were eliminated step by step since their regression coefficients are insignificant, which means these three first principal factors are not suitable to be used as their proxies. However, at least one of the proxies corresponding to "enterprise innovation activities", "education system" and "governance" is statistically significant according to the regression models in Table 5. These proxy variables include ENTRFGERD, TEENRL, POLSTAB, and REGULA. Hence, it may be better to substitute the 3 principal factors, ENTP1, EDUP1 and GOVP1 with these proxies. As for the proxy variables of relative income level, economic scale, dependency on foreign trade, and the natural endowments, the regression coefficients of PORGDP and TRTGDP are very robust in all the 5 models of Table 7, and are conformed to that in Table 5. Nevertheless, the regression results for RGDPPC and NRTGDP are different from that in Table 5. The regression results of RGDPPC in Table 5 are significant and robust, while those in Table 7 are insignificant. On the contrary, the regression coefficients of NRTGDP are totally insignificant in Table 5, while those in Table 7 shows fairly good significance, particularly in model 4 and model 5. To further improve the regression results of the panel data analysis, the proxies with good significance in the regression models in Table 5 together with 3 constructed first principal factors, INFRAP1, FINP1 and MARKP1, are selected as the explanatory variables. See Table 8. www.economics-ejournal.org 17 conomics Discussion Paper Table 8: Regression with principal factor and original proxy as explanatory Explanatory Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 C 0.025(0.883) -0.046(0.717) -0.144(0.443) -0.153(0.288) -0.294(0.049) -0.230(0.118) INFRAP1? 0.001(0.001) 0.001(0.000) 0.001(0.004) 0.001(0.002) 0.001(0.007) 0.001(0.014) FINP1? -0.001(0.000) -0.001(0.000) -0.001(0.000) -0.001(0.000) -0.001(0.000) -0.001(0.000) ENTRFGERD? 0.002(0.024) 0.002(0.015) 0.002(0.019) 0.002(0.015) 0.002(0.024) 0.002(0.048) TEENRL? -0.001(0.324) -0.001(0.420) – – – – SEENRL? – – 0.001(0.310) 0.001(0.244) 0.003(0.039) 0.001(0.205) MARKP1? -0.009(0.001) -0.009(0.001) -0.008(0.003) -0.009(0.001) -0.008(0.002) -0.007(0.006) GOVEFF? 0.019(0.482) – 0.011(0.670) – – – POLSTAB? 0.042(0.000) 0.043(0.000) 0.046(0.000) 0.048(0.000) 0.046(0.000) 0.037(0.002) REGULA? -0.081(0.002) -0.070(0.000) -0.070(0.007) -0.063(0.002) – – VOACCT? -0.050(0.058) -0.042(0.063) -0.055(0.040) -0.055(0.019) -0.059(0.015) RGDPPC? -0.016(0.616) – 0.001(0.983) – – – PORGDP? 0.068(0.000) 0.068(0.000) 0.064(0.000) 0.066(0.000) 0.068(0.000) 0.072(0.000) TRTGDP? 0.002(0.000) 0.002(0.000) 0.002(0.000) 0.002(0.000) 0.002(0.000) 0.002(0.000) NRTGDP? -0.003(0.105) -0.003(0.129) -0.004(0.089) -0.004(0.081) -0.003(0.227) -0.003(0.200) R-Sq 0.988 0.99 0.988 0.99 0.99 0.989 Adjusted R-Sq 0.985 0.988 0.985 0.988 0.988 0.987 D.W. 1.607 1.616 1.626 1.644 1.664 1.645 Cross section F 181.76(0.000) 212.26(0.000) 185.88(0.000) 217.37(0.000) 222.18(0.000) 223.07(0.000) In the 6 regression models in Table 8, the three principal factors, INFRAP1, FINP1, and MARKP1, still show good significance. And the coefficients are very robust across all the 11 models listed in Table 7 and Table 8. Regression coefficients of PORGDP, TRTGDP, and NRTGDP also show good robustness among all the 11 models. The coefficients of ENTRFGERD are significant and robust in the 12 regression models listed in Table 5 and Table 8. While the regression results of the two proxy variables for education system, SEENRL and TEENRL are very different from that in Table 5. The regression results for proxies of governance in Table 8 are much different from that in Table 5 as well with POLSTAB and REGUL changing their signs across the two tables. 3.5 Further analysis on the regression results Factors affecting NIS efficiency and their mechanisms Based on all the regression results in Table5, 7 and 8, it is clear how the NIS efficiency is affected by relevant influencing factors. And the affecting mechanisms behind can be further figured out with the guidance of innovation economic theory, NIS approach, and the new growth theory. Firstly, NIS efficiency is positively related to ICT infrastructure. Country with higher coverage of ICT infrastructure would have a better score. As a matter of fact, the diffusion of knowledge and information is highly dependent on ICT infrastructure. And the diffusion of knowledge and information is of great importance to innovation. After all, innovation is inherently a sort of knowledge-based activity. Secondly, the NIS efficiency of a country is affected by R&D and innovation activities of corporate sector. Proportion of enterprises in total R&D expenditure is positively related to efficiency score. From the view of innovation economics, www.economics-ejournal.org 18 conomics Discussion Paper enterprise is the most active and important element in an innovation system. The more enterprises were involved in innovation activities, the more efficient would the NIS be. Thirdly, larger economic scale and higher degree of openness would be helpful to form a more efficient NIS. Larger economic scale and higher dependent ratio means a bigger domestic and international market, and would facilitate the diffusion of knowledge and technology in a larger scope. It would be easier for innovative activities to gain the benefit of economy of scale and economy of scope in such a circumstance. Fourthly, the factors of financial structure, market circumstance, and governance are all relevant to NIS efficiency. However, the coefficients of the proxy variables are not very robust. As too many proxy variables were selected for these factors, the correlation among the proxies would affect the regression results more or less. Another reason is that the functioning mechanisms of these factors on NIS are not very direct. Good financial structure, market circumstance, and governance would only form a favorable external environment for innovation activities by reducing various kinds of transaction costs. Fifthly, the regression results for proxies of education system, natural endowments, and income level are not robust with the corresponding coefficients being significant in some models and insignificant in other models. In fact, these factors are only effective for NIS efficiency of some countries, and their impacts on NIS are not very direct. Education system would affect NIS efficiency through accumulation in human capital, which could be regarded as the source of innovativeness. And an economy highly dependent on natural resources might be short of incentives for innovation activities, which would in turn reduce its NIS efficiency. As for the income level, only a very limited number of countries can grasp the so called later comer advantages and step into a developed stage, which would be a positive impact on NIS efficiency. Further analyses on NIS efficiency for each country of BRICS Based on the previous econometric and mechanisms analyses, factors affecting NIS efficiency of each BRICS country can be further inferred with some basic data in Table 9. Many as the factors relevant to NIS efficiency, the decisive factors for each BRICS country are very different. For China, the high efficiency score and good ranking are mainly due to its huge economic scale, high proportion of enterprise R&D, and high dependency on foreign trade. In the year 2008, the GDP takes a proportion of 6.64% of the world total output, and the R&D financed by enterprises reaches to 72%, both much higher than those of the other BRICS countries. China also has a high dependency ratio on foreign trade of 66%, ranking the second among the BRICS www.economics-ejournal.org 19 conomics Discussion Paper countries. Moreover, China is still in a take-off stage according to income level per capita, and the industrialization is far from fulfilled. Thus, there are some later comer advantages for improving the innovation capacity. Nevertheless, China still has much room for progress in the fields of ICT infrastructure, education system, market circumstance and governance. Besides, China’s dependency ratio on natural endowments is still quite high comparing with other countries selected in this paper. All these elements would suppress micro-level innovation activities in the long run, and in turn impact NIS efficiency negatively. The fairly good efficiency score and ranking of India mainly comes from the huge later-comer advantages due to its backward developing stage. And the big economic scale is another element positively contributing to its NIS efficiency. As for the other relevant factors, the performance of India is very poor. Russia has great advantages in ICT infrastructure and education system among the BRICS countries, which is even comparable to that of most developed nations. However, Russia is highly dependent on its natural resources in recent years. And Russia government does not govern in a satisfactory manner. The badly ranking of Brazil can be attributed to its low proportion of enterprise R&D, low dependency on foreign trade, high dependency on natural resources, and the unsatisfying governance. Yet, Brazil still has relative advantages in ICT infrastructure and economic scale. As for the South Africa, the low efficiency of NIS is mainly due to the low coverage of ICT infrastructure, low participation of enterprises in R&D. Besides, South Africa has no advantages in economic scale and education system as well. However, the market circumstance, the governance, and the dependency ratio on foreign trade of South Africa are obviously better than those of the other BRICS countries. www.economics-ejournal.org 20 conomics Discussion Paper Table 9: Some of the data in 2008 for factors influencing NIS efficiency BR RU IN CN ZA US JP DE UK FR CA ITNET 37.5 32 4.5 22.5 8.6 75.8 75.2 78.1 78.2 70.4 75.3 TEL 21.5 31.6 3.3 25.7 9.1 50.8 37.9 62.2 54.1 56.2 54.8 ENTRRE 37 50 37 69 31 80 75 60 34 57 60 ENTRPGERD 40 63 34 73 58 73 78 69 62 63 54 ENTRFGERD 44 29 34 72 43 67 78 67 45 51 48 PORGDP 2.1 1.1 2 6.6 0.5 28.8 12.7 5.2 4.4 3.7 2.1 TRTGDP 27 53 52 62 74 31 35 89 61 56 69 NRTGDP 7.2 31 5.8 3.8 5.9 2.2 0.1 0.2 2.4 0.1 8 RGDPPC 1.6 2.4 0.5 0.9 1.6 7.2 5.2 5.6 5.6 5.1 5.9 TEENRL 34 77 13 23 32 83 58 60 57 55 70 SEENRL 82 85 60 76 73 88 98 92 93 98 91 TAXRATE 69 48 69 80 34 47 55 51 35 65 45 GOVEFF 0.1 -0.3 0 0.2 0.7 1.5 1.4 1.4 1.6 1.5 1.8 POLSTAB -0.1 -0.6 -0.9 -0.4 0.2 0.5 0.9 1 0.5 0.6 1 REGULA 0.1 -0.5 -0.3 -0.1 0.5 1.5 1.1 1.4 1.7 1.2 1.6 IT FI SE DK CH NL NO AT BE AU KR ITNET 44.4 83.5 89 84.5 70.8 80.2 68.9 87.9 90.5 72.9 70.5 TEL 35.5 31.1 57.7 45.3 43.7 43.9 63.1 44.5 39.8 39.4 41.6 ENTRRE 38 59 69 66 29 77 41 49 51 63 47 ENTRPGERD 53 74 74 70 61 75 74 50 54 71 68 ENTRFGERD 45 70 61 61 61 73 68 49 46 46 61 PORGDP 2.9 0.4 0.8 0.4 1.3 1.9 0.7 1.1 0.5 0.6 0.7 TRTGDP 58 90 100 107 41 107 102 145 77 113 171 NRTGDP 0.3 0.9 1.1 3.6 8.6 0 0 2.7 21.8 0.5 0 RGDPPC 4.7 5.6 5.7 5.6 5.7 4.2 6.3 4.5 7 4.5 4.1 TEENRL 67 94 71 78 77 98 49 61 73 55 63 SEENRL 95 96 99 90 88 95 85 88 96 99 86 TAXRATE 73 48 55 30 50 34 29 39 42 55 58 GOVEFF 0.4 2 1.9 2.1 1.8 1.1 1.9 1.7 1.8 1.6 1.2 POLSTAB 0.6 1.4 1.1 1 1 0.4 1.2 0.9 1.3 1.3 0.7 REGULA 0.9 1.6 1.6 1.9 1.7 0.7 1.6 1.7 1.4 1.6 1.3 4 Concluding remarks The previous parts of this paper calculate the relative efficiency of NIS for 22 countries including the BRICS from 2000 to 2008 considering the national level innovation inputs and outputs with DEA method. Taking the efficiency scores as the explained variable, factors affecting NIS efficiency are further analyzed with econometric and statistical tools. The efficiency calculation and the empirical test results can be summarized: (1) The BRICS are very different in their relative efficiency of NIS. Russia, India and China have relatively high efficiency score and good ranking, while www.economics-ejournal.org 21 conomics Discussion Paper Brazil and South Africa are not, ranking at the bottom among the 22 selected countries. (2) Influencing factors of NIS efficiency involve a lot of elements, including the ICT infrastructure, enterprise R&D activities, economic scale, economic openness, financial structure, market circumstance, governance, education system, natural endowments. This is conformed to the relevant arguments of NIS Approach and the New Growth Theory. (3) Enterprise is the most active and important actor for innovation and enterprises innovation activities are of key importance to the NIS. The more enterprises involved in R&D activities, the higher would the NIS efficiency be. Elements including financial structure, market circumstance, and governance level form the external environments for innovation activities, which would affect NIS efficiency indirectly. (4) ICT infrastructure, economic scale, and openness decide diffusion speed and scope of knowledge, and in turn affect NIS efficiency. Furthermore, economic scale and degree of openness decide the scale of domestic and international market for enterprises. The economy of scale and economy of scope are much easier to be realized in a bigger market, which would influence the NIS efficiency indirectly as well. (5) The decisive factors for NIS efficiency of each BRICS are very different. However, the BRICS still have some characters in common, particularly the low governance and fairly high dependency on natural resources, which was decided by their developing stage and extensive developing patterns. In modern history, only very few economies, e.g. Japan and South Korea, have caught up with and leaped into developed nations successfully. In the coming future, the BRICS should endeavor a transition from factors-driven to innovationdriven pattern in order to improve their competitiveness substantially and avoid "middle-income trap". Governments of each country should improve its governance and create a sound external environment for enterprise innovation. China is implementing its "12th Five Year Planning" for national economy and social development from 2011 to 2015. To accelerate the process of constructing innovative nation, and transform the economic pattern essentially, governments in different levels should dedicate to improve capability in social administration, enhance the expenses on infrastructure and education, and create a more comfortable market circumstance. References [1] P. Aghion and P. Howitt. Endogenous Growth Theory. The MIT Press, Cambridge, Massachusetts., 1998. www.economics-ejournal.org 22 conomics Discussion Paper [2] José Cassiolato and Bengt-Aake Lundvall. Brics project: Comparative and summary report on brics national innovation system. Research project report, Jointly by IDRC, CRDI, RedeSist, Globelics and UFRJ, Brazil, 2010. [3] Barbara Casu and Philip Molyneux. A comparative study of efficiency in european banking. Applied Economics, 35:1865–1876, 2003. [4] A. Charnes, W.W. Cooper, and E. Rhodes. Measuring the efficiency of decision making units. European Journal of Operational Research, 2:429– 444, 1978. [5] M. J. Farrell. The measurement of productive efficiency. Journal of the Royal Statistical Society. Series A (General), 120(3):253–290, 1957. [6] Stephen Feinson. National innovation systems overview and country cases. Research project report, Jointly by IDRC, CRDI, RedeSist, Globelics and UFRJ, 2003. [7] Chris Freeman. Continental, national and sub-national innovation systemscomplementarity and economic growth. Research Policy, 31:191–211, 2002. [8] Jeffery L. Furman and Richard Hayes. Catching up or standing still? national innovative productivity among ’follower’ ountries, 1978-1999. Research Policy, 33:1329–1354, 2004. [9] Jeffery L. Furman, Michael E. Porter, and Scott Stern. The determinants of national innovative capacity. Research Policy, 31:899–933, 2002. [10] Jian-Cheng Guan, Richard C.M. Yam, C K Mok, and Ning Ma. A study of the relationship between competitiveness and technological innovation capability based on dea models. European Journal of Operational Research, 170:971–986, 2006. [11] Ayoe Hoff. Second stage dea: Comparison of approaches for modeling the dea score. European Journal of Operational Research, 181:425–435, 2007. [12] Mei-Chih Hu and John A. Mathews. National innovative capacity in east asia. Research Policy, 34:1322–1349, 2005. [13] Mei-Chih Hu and John A. Mathews. China’s national innovative capacity. Research Policy, 37:1465–1479, 2008. [14] X. Liu and S. White. Comparing innovation systems: a framework and application to china’s transitional context. Research Policy, 30:1091–1114, 2001. www.economics-ejournal.org 23 conomics Discussion Paper [15] W. Nasierowski and F. J. Arcelus. On the efficiency of national innovation systems. Socio-Economic Planning Sciences, 37:215–234, 2003. [16] Richard Nelson. Economic development from the perspective of evolutionary economic theory. Oxford Development Studies, 36:9–21, 2008. [17] P. Patel and K. Pavitt. National innovation systems: why they are important, and how they might be measured and compared. Economics of Innovation and New Technology, 3(1):77–95, 1994. [18] V.J. Tomas, Seema Sharma, and Sudhir K. Jain. Using patents and publications to assess r&d efficiency in the states of the usa. World Patent Information, 33:4–10, 2011. [19] Eduardo B. Viotti. National learning systems: A new approach on technological change in late industrializing economies and evidences from the cases of brazil and south korea. Technological Forecasting & Social Change, 69:653–680, 2002. (Vol.1)". www.economics-ejournal.org 24 Please note: You are most sincerely encouraged to participate in the open assessment of this discussion paper. You can do so by either recommending the paper or by posting your comments. Please go to: http://www.economics-ejournal.org/economics/discussionpapers/2011-52 The Editor © Author(s) 2011. Licensed under a Creative Commons License - Attribution-NonCommercial 2.0 Germany